A Preliminary Investigation on the Usage of Quantum Approximate Optimization Algorithms for Test Case Selection

TL;DR

Regression testing faces scalability limits, motivating test case selection using quantum optimization. The authors propose QAOA-TCS, formulating a multi-objective QUBO solved by QAOA with clustering to fit hardware, and evaluate it in an ideal StatevectorSimulator against classical baselines and SelectQA. Results show QAOA-TCS achieves superior effectiveness (larger Pareto fronts) while maintaining comparable efficiency, suggesting potential gains when deployed on real quantum devices. The work lays groundwork for noisy-environment evaluation and hardware experiments.

Abstract

Regression testing is key in verifying that software works correctly after changes. However, running the entire regression test suite can be impractical and expensive, especially for large-scale systems. Test suite optimization methods are highly effective but often become infeasible due to their high computational demands. In previous work, Trovato et al. proposed SelectQA, an approach based on quantum annealing that outperforms the traditional state-of-the-art methods, i.e., Additional Greedy and DIV-GA, in efficiency. This work envisions the usage of Quantum Approximate Optimization Algorithms (QAOAs) for test case selection by proposing QAOA-TCS. QAOAs merge the potential of gate-based quantum machines with the optimization capabilities of the adiabatic evolution. To prove the effectiveness of QAOAs for test case selection, we preliminarily investigate QAOA-TCS leveraging an ideal environment simulation before evaluating it on real quantum machines. Our results show that QAOAs perform better than the baseline algorithms in effectiveness while being comparable to SelectQA in terms of efficiency. These results encourage us to continue our experimentation with noisy environment simulations and real quantum machines.